HDF5 1.8.14 release candidate (pre1) is available for testing

Dear HDF5 users,

A pre-release candidate of HDF5 1.8.14 for UNIX and Windows is available for testing and can be downloaded from the following location:

ftp://ftp.hdfgroup.uiuc.edu/pub/outgoing/hdf5/hdf5-1.8.14-pre1/

There are two release source files in the directory:
hdf5-1.8.14-pre1.tar.gz is for Unix platforms
hdf5-1.8.14-pre1.zip is for Windows platforms

We will greatly appreciate if you test this pre-release on the platform(s) of your choice and report any problems found to this list as soon as possible. We have about one week to address any issues.

Notice to the autoconf Unix users: The default configuration for this release candidate is “production”.
If you wants more debugging code on, use “./configure —disable-production …”.

The list of the features, bug fixes, and platforms tested by The HDF Group can be found in the

ftp://ftp.hdfgroup.uiuc.edu/pub/outgoing/hdf5/hdf5-1.8.14-pre1/hdf5-1.8.14-pre1-RELEASE.txt

Highlights:
* Chunk Fill Writes Changed to Collective for better performance
* Initialization of Object IDs by setting “id” initially to H5I_INVALID_HID
* Removed SVN URLs from cmake files
* Fixed a bug in Cmake packaging for a missing Fortran module.
* Shared Library Version Numbers Changed (soname) due to interface changes.
* Configure Setting for Mac OSX has proper default values for PROD_XXX, DEBUG_XXX, and PROFILE_XXX flags
* CMake ConfigureChecks added two missing include files.
* Bison was upgraded to 2.7, and Flex was upgraded to 2.5.37, m4 upgraded to 1.4.17
* Autotools: Modified configure to add an entry at the beginning of AM_LDFLAGS for the hdf5 install directory.
* Configure flags error if REAL is same as DOUBLE PRECISION and both Fortran and HL library are enabled.
* Fixed Identifier Management Code which failed to reuse identifiers what have been freed.
* Removal of DllMain() from Static Windows Builds.
* Enforce Constraint on page_size Parameter in H5Pset_core_write_tracking().
* H5Ldelete_by_idx() fails on non-existent group name fixed.
* Bug in Test When Building Parallel HDF5 on PVFS2 fixed.
* MPI-IO Driver Tried to Allocate Space for Zero-length Dataset error fixed.
* SIZEOF Replaced by C_SIZEOF and STORAGE_SIZE.
* Non-functional API: h5pget_fill_value_f fixed.
* Interoperability with C HDF5: H5Literate and h5literate_f failure fixed.
* Interoperability with C HDF5: Constant INTEGER Parameters with the H5FD Interface failure fixed.
* Removed H5F_ACC_CREAT from C++ API.
* Seg Faults in H5TBread_field_name and H5TBread_field_name_f fixed.
* Possible Buffer Overflow in High-level (HL) APIs fixed.
* H5LTdtype_to_text was changed to truncate the string if the user buffer is too small.
* Visual Studio 2013 support is added for Windows platforms.

Please don't be discouraged if you don't see your patch in this release. Some patches came too late or need rework.

Happy testing and we look forward to hearing from you!

Thank you!

Albert Cheng
THG Staff

···

~
~
~

on cygwin 64 bit no regression compared to 1.8.13
same failures as before

$ grep FAIL *.chklog
cache_api.chklog:Testing MDC/FILE related API calls *FAILED*
cache_api.chklog:*FAILED*
dt_arith.chklog:Testing query functions of compiler conversion *FAILED*
dt_arith.chklog:***** 1 FAILURE! *****
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB3 filter for group *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB3 filter for group *FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB1 filter *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing opening groups with DYNLIB3 filter *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:***** 6 PLUGIN TESTS FAILED! *****
testerror.sh.chklog:Testing err_compat *FAILED*
testerror.sh.chklog: ! Testing error API H5Eset/get_auto *FAILED*
testerror.sh.chklog: ! *FAILED*
testerror.sh.chklog: ! ***** ERROR TEST FAILED! *****
testerror.sh.chklog:Testing error_test *FAILED*

Regards
Marco

···

On 10/29/2014 10:40 PM, Albert Cheng wrote:

Dear HDF5 users,

A pre-release candidate of HDF5 1.8.14 for UNIX and Windows is available for testing and can be downloaded from the following location:

We will greatly appreciate if you test this pre-release on the platform(s) of your choice and report any problems found to this list as soon as possible. We have about one week to address any issues.

The dynamic loading on cygwin is in the issue database because of cygwin vs *nix library naming issues.

Allen

···

________________________________________
From: Hdf-forum <hdf-forum-bounces@lists.hdfgroup.org> on behalf of Marco Atzeri <marco.atzeri@gmail.com>
Sent: Thursday, October 30, 2014 2:13 AM
To: hdf-forum@lists.hdfgroup.org
Subject: Re: [Hdf-forum] HDF5 1.8.14 release candidate (pre1) is available for testing

On 10/29/2014 10:40 PM, Albert Cheng wrote:

Dear HDF5 users,

A pre-release candidate of HDF5 1.8.14 for UNIX and Windows is available for testing and can be downloaded from the following location:

We will greatly appreciate if you test this pre-release on the platform(s) of your choice and report any problems found to this list as soon as possible. We have about one week to address any issues.

on cygwin 64 bit no regression compared to 1.8.13
same failures as before

$ grep FAIL *.chklog
cache_api.chklog:Testing MDC/FILE related API calls
                *FAILED*
cache_api.chklog:*FAILED*
dt_arith.chklog:Testing query functions of compiler conversion
               *FAILED*
dt_arith.chklog:***** 1 FAILURE! *****
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB3 filter for group
                     *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB3 filter for group
                     *FAILED*
test_plugin.sh.chklog:Testing Testing DYNLIB1 filter
                     *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:Testing Testing opening groups with DYNLIB3 filter
                    *FAILED*
test_plugin.sh.chklog:*FAILED*
test_plugin.sh.chklog:***** 6 PLUGIN TESTS FAILED! *****
testerror.sh.chklog:Testing err_compat
                   *FAILED*
testerror.sh.chklog: ! Testing error API H5Eset/get_auto
                         *FAILED*
testerror.sh.chklog: ! *FAILED*
testerror.sh.chklog: ! ***** ERROR TEST FAILED! *****
testerror.sh.chklog:Testing error_test
                   *FAILED*

Regards
Marco

_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Hardcoded filename in the tests ?

I see that in test_plugin.sh, but I assume there are other points as

Testing DYNLIB1 filter
*FAILED*
HDF5-DIAG: Error detected in HDF5 (1.8.14-pre1) thread 0:
   #000: /pub/devel/hdf5/hdf5-1.8.14pre1-1.x86_64/src/hdf5-1.8.14-pre1/src/H5Pocpl.c line 756 in H5Pset_filter(): fail ed to call private function
     major: Property lists
     minor: Can't set value
   #001: /pub/devel/hdf5/hdf5-1.8.14pre1-1.x86_64/src/hdf5-1.8.14-pre1/src/H5Pocpl.c line 818 in H5P__set_filter(): fa iled to load dynamically loaded plugin
     major: Data filters
     minor: Unable to load metadata into cache

Regards
Marco

···

On 10/30/2014 3:55 PM, Allen Byrne wrote:

The dynamic loading on cygwin is in the issue database because of cygwin vs *nix library naming issues.

Allen

No in the library. The search process for a dynamic library in the "not"
windows section was written with *.so as the pattern.

Allen

···

On Thursday, October 30, 2014 05:40:24 PM Marco Atzeri wrote:

On 10/30/2014 3:55 PM, Allen Byrne wrote:
> The dynamic loading on cygwin is in the issue database because of cygwin
> vs *nix library naming issues.
>
> Allen

Hardcoded filename in the tests ?

I see that in test_plugin.sh, but I assume there are other points as

Testing DYNLIB1 filter
*FAILED*
HDF5-DIAG: Error detected in HDF5 (1.8.14-pre1) thread 0:
   #000:
/pub/devel/hdf5/hdf5-1.8.14pre1-1.x86_64/src/hdf5-1.8.14-pre1/src/H5Pocpl.c
line 756 in H5Pset_filter(): fail ed to call private function
     major: Property lists
     minor: Can't set value
   #001:
/pub/devel/hdf5/hdf5-1.8.14pre1-1.x86_64/src/hdf5-1.8.14-pre1/src/H5Pocpl.c
line 818 in H5P__set_filter(): fa iled to load dynamically
loaded plugin
     major: Data filters
     minor: Unable to load metadata into cache

Regards
Marco

ok, found it

adjusting for the right assumptions:

···

On 10/30/2014 8:47 PM, Allen Byrne wrote:

No in the library. The search process for a dynamic library in the "not"
windows section was written with *.so as the pattern.

Allen

------------------------------------------------------------------------------
Testing test_plugin.sh
Finished testing test_plugin.sh

test_plugin.sh Test Log

Testing with old file format:
Testing deflate filter
Testing filters (setup) PASSED
Testing filters (uninitialized read) PASSED
Testing filters (write) PASSED
Testing filters (read) PASSED
Testing filters (modify) PASSED
Testing filters (re-open) PASSED
Testing filters (partial I/O) PASSED
Testing DYNLIB1 filter
Testing filters (setup) PASSED
Testing filters (uninitialized read) PASSED
Testing filters (write) PASSED
Testing filters (read) PASSED
Testing filters (modify) PASSED
Testing filters (re-open) PASSED
Testing filters (partial I/O) PASSED
Testing DYNLIB2 filter
Testing filters (setup) PASSED
Testing filters (uninitialized read) PASSED
Testing filters (write) PASSED
Testing filters (read) PASSED
Testing filters (modify) PASSED
Testing filters (re-open) PASSED
Testing filters (partial I/O) PASSED
Testing Testing DYNLIB3 filter for group PASSED
------------------------------------------------------------------------------

Marco

The attached patch correct the assumption and the test.

There is still missing point;
libdynlib1.la libdynlib2.la libdynlib3.la should be declared noinst
to avoid their installation

hdf5-1.8.14pre1-1.src.patch (1.48 KB)

···

On 10/30/2014 10:10 PM, Marco Atzeri wrote:

On 10/30/2014 8:47 PM, Allen Byrne wrote:

No in the library. The search process for a dynamic library in the "not"
windows section was written with *.so as the pattern.

Allen

ok, found it

adjusting for the right assumptions:
------------------------------------------------------------------------------